Identifying and counting commercial solar installations in South Africa

Tsepang Polaki , MH van Staden

Partner: solar-geography

Year: 2022

Abstract: Solar power is a major player in the pursuit of clean, environmentally friendly energy production. This project seeks to help facilitate the transition by building a model capable of detecting solar panel installations across the Tshwane municipality. A sample of satellite images and the associated masks have been provided and are used to train the model. The Region Based Convolutional Neural Networks (RCNN) model produced managed to detect up to 50% of solar panels in suburbs in the Old East area of Pretoria. This culminated in an interactive web app, capable of showing the location, detected number of panels, estimated energy produced and sample images with solar panels detected. Thus giving the municipality valuable insight as to where and to what extent solar energy has been embraced or neglected.

Presentation Video